Statistics for 2007 Semester 2

This Page shows some statistics from the 2007 Semester 2 allocations

Preference outcomes

Please see the 2007 semester 1 stats for an explanation of this table

Clash Statistics

Note There are 115802 allocation made by OLCR. These stats should be read with that in mind.
I have blanked out the links to example students.

Unfixable clashes are ones where the classes only offer one option and those options clash
eg Unit A Lecture no repeats is at the same time as Unit B Lecture no repeats. Only a change in Lecture time or student enrolment will resolve this clash.
OLCR allocates students into the times and venues determined by Timetables. It can not resolve these clashes.

Timetables have a very difficult problem creating a timetable that works for all students.
Of the 16,000 odd students I analysed 11,000 of them had differing unit enrolments. That’s 11,000 individual timetable needs that also needs to allow for lecturer, tutors and room availability.

Please note that OLCR has a test system that all OLCR admins have access too. This test system does nightly test allocations.
All of these clashes are visable to both OLCR admins and Timetable staff long before the actual allocation occurs. The OLCR admin system gives a full breakdown on which units are involved and how large the clash problems are in each case. It also has reports that show where every student enrolled in the unit is during the week in 1hr blocks to aid in locating suitable timetable changes that might resolve these clashes.
Even with these tools. Resolving these Unfixable clashes is a complicated and difficult task.

Maybe Clashes are clashes that may be resolvable or may be unfixable. They are reported as maybe clashes because the clash report page does not attempt a detailed analysis.
In 2006 I did a detailed manual analysis of the semester 1 OLCR run. The results of that analysis would apply to this allocation.

The results where that 50% of the maybe clashes where more complicated unfixable classes.
eg Unit A has a lecture. Unit B has a lecture. Unit C has two Tutes, one at the same time as the unit A lecture and the other at the same time as the unit C lecture. A group of students are enrolled in all three units. There is no solution for the two tutes since either option has a clash.

I also include clashes that indirectly result from class size limits in this catergory.
An example of this would be: Unit A has 1 Lecture, Unit B has a Lecture which has a repeat Lecture (ie 2 options, attend one of the two lectures. The content in both lectures is the same). The Lecture in Unit A is at the same time as one of the lectures in Unit B. The unit coordinator for Unit B sets the class sizes to half the enrolled student count in an effort to even up the classes. If too many students are enrolled in both Unit A and B then these students won't all be able to fit into the non-clashing lecture in Unit B and OLCR is force to allocate them a clash.

The remaining 50% where caused by:

1) Having two allocations. In the second allocation it can not move allocations made in the 1st allocation or by students and unit admins. As a workaround for this the 1st allocation actually allocates all the units and then deletes all but the 1st allocation units. These leaves "holes" for the second allocation, but often students move into these holes blocking the 2nd allocation.

2) Incorrectly or insufficently configure units. Some units have special classes that are scheduled into the same timeslot as other normal classes. ie The lecture in week 6 will actually be a test. The unit is configured with both the lecture and the test running all of semester.

3) Students delibrately choosing to clash their classes to avoid having to attend classes a inconvent times.
eg Rather that go to the tute at 5pm on friday, I'll go to the one on Monday at 11 and skip my lecture that is at that time. Lectures content is often available online or from other students.

Note a test run of the OLCR clash and preference optimisers where run with the freedom to change any allocation and after incorrectly unit configurations where removed. This resulted in the removal of this 50% of maybe clashes. (ie the problem isn't with OLCR)

Clashes come in pairs. The above #clashes have been divided by two to allow for this. ie Only one of the two allocations is viewed as at fault, not both.